[P4-discuss] table entry population

Leo Alterman leo at barefootnetworks.com
Thu Apr 23 21:21:50 EDT 2015


Hi Brian -
I believe the intention was to do things like learning through the
'generate_digest' primitive, which leaves it open to the target what
endpoints are there to receive the digest and what is actually done with
them (eg, whether it'll end up at the CPU or is an entirely fast-path
operation).

In the posted example at
https://github.com/p4lang/p4factory/blob/master/targets/dc_example/p4src/l2.p4
line 295, learning is triggered by sending a digest to the '
MAC_LEARN_RECIEVER' endpoint. Depending on the target this might or might
not result in CPU intervention. It's probably a good idea to have different
receivers to differentiate between CPU-based vs fast-path learning, but
that's ultimately the decision of whoever is implementing the compiler
you're using.

~Leo

On Thu, Apr 23, 2015 at 6:03 PM, Brian Brooks <brooks.brian at gmail.com>
wrote:

> Is it possible to do L2 src mac learning in P4 without relying on a
> "copy_to_cpu" action and having the run-time system to insert the mac?
>
> _______________________________________________
> P4-discuss mailing list
> P4-discuss at p4.org
> http://p4.org/mailman/listinfo/p4-discuss_p4.org
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.p4.org/pipermail/p4-discuss_lists.p4.org/attachments/20150423/8af7ed99/attachment-0001.html>


More information about the P4-discuss mailing list